Home Surveillance System (HSS) - UPDATE and UPGRADE

Share Your Projects
21

If you are happy just re-using the IP CAM10 to 14 tabs you can leave the hss_config as is and just change the url in the inject node. But if you want additional cameras you need to duplicate complete tabs and also add those to the hss_config. If you in addition would like to have another label or rotate the view, you also have to modify the hss_config

Once done, you can set the inject node to inject once at start so you always have your camera views ready for viewing and recording

2 Likes
22

Any reason not to nuke my existing HSS install in node-red and then install the new version starting at step 3, assuming the nginx I installed with the initial version would still be valid?

I had health issues in 2022 and am finally back to nearly normal. I had put the Pi4B-2GB running it on a shelf until now. I'd mucked around with it trying to learn advanced node-red features, of which this seems to be a showcase, so I don't think it would now be a good base system to upgrade.

My python based DVR add-on with yolo8 verification and minimal node-red user interface has performed admirably all this time, so I've no plans on really using it, but I'd like to play with it and perhaps recommend it to others who find my system "overwhelming" to set up.

Is anyone running it on a Pi5 and 64-bit Bookworm? I ask as I've just installed Bookworm 64-bit on a Pi4B-4GB to evaluate yolo11, a Pi5 is recommended:
https://docs.ultralytics.com/guides/raspberry-pi/

Apparently the yolo11n model exported to OpenVINO format can use the PI5 GPU and get ~12 fps with mAP50-95(B) of ~0.60

This could bring the Pi5 into play for my system since the Nvidia 1060 in the laptop I've been running for a couple of years only gets about 15 fps albeit with the much larger yolo8x model with an mAP of about 0.54

Are there any docs or threads about adding an object detector to the motion detected images in HSS? I was just looking into this when health went bad for me. From what little I could keep up with it, from Bart's threads it seems pure node-red AI object detection didn't seem to pan out. A lot can be done reasonable effort using MQTT and an exec node launching an independent Python program at startup.